Client: Leading European university hospital
Industry: Medtech & Life Science
Solutions Area: AI, Medical Image Analysis
From manual pixels to AI‑driven support
Before treating patients with brain tumours, specialists need to manually outline the tumour volume, image by image. It is meticulous, essential and takes many hours per patient.
When the same work is also required for large research studies, it quickly becomes a bottleneck. The number of patients that can be included is limited, analyses take longer, and the path from hypothesis to published research becomes unnecessarily long.
Here, the university hospital saw a clear opportunity: can AI handle the most repetitive part of the job, so that clinicians can spend their time where it creates the most value?
AI models that see what the clinician needs
HiQ took an end‑to‑end approach to the technical solution, from data handling to model evaluation, in close collaboration with researchers and clinicians.
Together we defined datasets, annotations and quality metrics so that every test could be tracked and compared. Based on this, we built several advanced AI models that automatically identify and delineate tumours in brain MRI.
Each model was optimised for different strengths, such as sensitivity to small changes in brain tissue, robustness to noise and varying image quality, and stability across different patient cases and protocols.
When one of the models clearly outperformed the others, the team took the next step: combining several networks in an ensemble. By allowing the models to “vote” on the tumour boundaries, we could further improve accuracy and achieve more robust segmentation.
Results that move the field forward
The evaluation showed that all models performed well, with a clear top candidate that achieved even greater precision than any individual model.
At the same time, the bar for clinical implementation in neuroradiology and oncology is very high, and rightly so. Accuracy is not yet at the level required for full-scale use in everyday clinical practice, but the direction is clear. The university hospital has been provided with a concrete set of models and a codebase to further develop, and the research group has a structured method for evaluating AI-based segmentation against established practices.
The next steps towards clinical decision support have been defined, with a focus on larger datasets, broader validation and integration into existing systems.
A new standard for the future of healthcare with advanced image analysis
This is not a lab experiment tucked away in a corner of the organisation. It is practical proof of how AI can take a role in healthcare’s most sensitive decision flows – without taking control away from the clinician.
When segmentation is automated and standardised, healthcare can:
- free up valuable specialist time
- increase the number of patients in research studies
- gain more consistent and comparable volumes over time
- shorten the path from imaging data to analysis and decision.
For the patient, that means faster answers, more precise treatments and care where advanced technology actually makes a difference in the encounter – not just in presentation slides.
AI that lets experts focus on what really matters
The project shows what happens when deep medical expertise meets HiQ’s strength in software, data and AI. We do not build “AI features” just to talk about them – we build tools that step into real‑world workflows, hold up to scrutiny and can be scaled.
When a leading university hospital dares to bring AI into the analysis of brain tumours – and does so together with us – it sets the tone for what the next generation of MedTech solutions will look like.
Curious how you can free up time and increase precision with AI? Get in touch!
